No-Mortality Liver Resection for Hepatocellular Carcinoma in Cirrhotic and Noncirrhotic Patients: Title and subTitle BreakIs There a Way? A Prospective Analysis of Our Approach FREE

HEPATOCELLULAR carinoma (HCC) is one of the most frequently occurring tumors in humans, with more than 250,000 new cases per year diagnosed in the world.¹ It generally grows in a cirrhotic liver, and its incidence in patients with this condition is 3%.² This strict association has led some authors to consider the cirrhotic liver to be a preneoplastic condition.³ These factors, and the fact that the liver cirrhosis itself significantly reduces life expectancy, contribute to the ongoing controversy about the optimal therapeutic approaches to HCCs. The double aims of curative therapy for the tumor and the preservation of as much functioning liver parenchyma as possible are the priorities of any treatment modality for HCCs in patients with liver cirrhosis.

Historically, liver resection has been considered the treatment of choice, with homogeneous long-term survival rates achieved with different surgical series.^{4 - 9} In line with these results, Makuuchi et al¹⁰ reported 5-year survival of 44% for patients with single HCCs of less than 5 cm in diameter. Despite these favorable results, the advent of other effective treatments has called into question the real effectiveness of surgery for HCCs in patients with liver cirrhosis due to the following weak points of surgery: a low resectability rate, high morbidity, and a mortality rate that is still not negligible.

Since the 1980s, we have developed a precise policy limiting the unbalancing impact of those factors on the cost-benefit ratio of surgery. To validate the adequacy and safety of our policy, we performed a prospective analysis of the outcomes of patients with HCC who were referred to our department and underwent liver resection.

The terminology we have used for the surgical anatomy of the liver is based on the classification of Couinaud.¹¹

Resectability was assessed and the surgical procedure was selected by following an algorithm based on 3 parameters: the presence or absence of ascites, total serum bilirubin level, and indocyanine green retention rate at 15 minutes (ICG15) (Figure 1).¹² The feasibility of surgical resection was also determined on the basis of the liver volume, which was calculated, as described previously,¹³ by measuring and analyzing computed tomographic images. If right and extended right hepatectomy or left trisectoriectomy were planned and the volume of the remnant liver was expected to be below 40% of the entire hepatic volume or the ICG15 value was 10% to 20%, preoperative portal venous branch embolization was performed to reduce the risk for postoperative liver failure. If the selected surgical procedure guaranteed radical removal of the tumor, the patient underwent surgery.

The anatomical features of the liver and clinical stage of the neoplasm were defined after performing ultrasonography (US), plain and enhanced spiral computed tomographic scannings, plain and enhanced (with gadolinium) magnetic resonance imaging, and angiography using iodized oil (Lipiodol; André-Gelbe Laboratoires, Aulnay-sous-Bois, France) as contrast medium followed by computed tomography.

Preoperative bed rest was scheduled for each patient, and for this purpose a central venous catheter was introduced 1 week before the operation. If the level of both or at least 1 of the serum aminotransferases was above 100 IU/L, ursodesoxycholic acid as first choice, Shosaiko-to (Shosaiko-to; Tumura, Tokyo, Japan) as second choice, or glycyrrhizin (Stronger Neo-Minophagen C; Minophagen Pharmaceutical Co, Tokyo) as third choice was administered.

The following 3 types of incision were made: J shaped, upper median, and inverted T shaped. The operative field waskept open using special retractors (Kent retractors; Takasago, Tokyo) throughout the operation. If the lesion was located in segments 2, 3, and/or 4, an inverted T-shaped or upper-median abdominal incision was considered adequate. For lesions located in the right segments, the J-shaped incision was chosen, and it was extended to the ninth intercostal space. If the tumor to be resected was located in segments 7 or 8, or the right side of the liver was atrophic and elevated relative to the costal arch, the thoracic cavity was opened.

Thorough exploration of the liver with intraoperative US was performed after partial mobilization of the organ itself, sectioning the falciform and triangular ligaments and pulling the liver caudally to better expose its diaphragmatic surface. Definition of the area to be resected, monitoring of the dissection line, control of the complete removal of the tumor, and all maneuvers, including puncture, were performed under intraoperative US guidance.

Care was taken to avoid accidental bleeding, and blood transfusions were administered only if strictly necessary. To obtain a bloodless operative field, slow and accurate dissection was performed using mainly electrocautery to dissect and to cut. Every vessel was ligated with thin (2/3-0) silk sutures, and liver dissection was accomplished using Péan forceps. To avoid postoperative bile leakage after resection, intraoperative cholangiographic monitoring was also performed.

The goal of minimal blood loss was achieved by performing liver transection under warm ischemia, using the following techniques: the Pringle maneuver¹⁴ with clamping for 15 minutes at 5-minute intervals was performed for patients undergoing hepatectomy, extended hepatectomy, or segmentectomy (segments 5, 6, 2, and 3), and hemihepatic selective vascular occlusion was performed for all the other patients.¹⁴ With the latter technique, the hepatic hilum generally was dissected, and the clamping periods lasted 30 minutes at 5-minute intervals.

Drains were always left in the peritoneal cavity.

To reduce blood loss and avoid blood transfusions, volemia and the level of anesthesia were controlled. Blood loss and ascites production during the operation were balanced by infusing 10% to 20% more fresh frozen plasma (FFP) than the volume of blood lost. Intraoperative blood transfusions were given only if the hematocrit value was below 0.30. The amount of sodium infused was restricted, because the sodium concentration of FFP was high. To ensure a total fluid volume of 4 to 4.5 mL/kg per hour was infused, equal volumes of crystalloid (Ringer lactate) and 5% glucose solution were administered.

The level of anesthesia was maintained by general and epidural anesthesia, thereby reducing the quantity of inhalation agents and intravenous drugs. A muscle relaxant was also administered, and the respiratory tidal volume was reduced to about 60% just before starting liver dissection to reduce the thoracic and right atrial pressures and, consequently, the back bleeding from the hepatic veins and/or their tributaries. Hydrocortisone sodium phosphate (100 mg) was injected intravenously before starting vascular occlusion to protect the liver during warm ischemia.

The main aim during the first few postoperative days was to restore the liver function and avoid hepatic failure. The total serum protein level was maintained at about 65 g/L by administering about 15 mL/kg per day of FFP. The alkalosis induced by infusing such a large amount of FFP was balanced by administering a chloride-rich amino acid solution, which was switched to an acid-base–balanced solution when the FFP transfusion was stopped. The high sodium level of the FFP was balanced by infusing a sodium-free crystalloid. The patients with cirrhotic livers received a total fluid infusion of about 40 to 45 mL/kg per day. Glucose solution was given for the energy (calories), hydrocortisone was also given intravenously for 4 days postoperatively, and short-term antibiotic prophylaxis, histamine blockers, and low-molecular-weight heparin sodium were also administered. Appropriate oral intake was restored as soon as possible, because liver function improves better in patients receiving enteric rather than parenteral nutrition.

If the postoperative course was uneventful, the gastric tube and chest drain (if placed) were removed as soon as possible (during the first 2-4 postoperative days), whereas the abdominal drains were left for at least 1 week. At first, blood discharge from the drain was monitored carefully. If it exceeded 100 mL/h, an emergency laparotomy was performed. Blood transfusions were administered only to patients with hematocrit values below 0.20. During the second postoperative week, providing no bile leakage was detected, the abdominal drains were gradually removed. If bile was detected in the drain discharge, the tube was left until the fistula closed spontaneously, taking care to avoid compression of the liver surface with the tip of the drain. When pure bile without bacterial contamination or debris was drained, the drain was progressively reduced in caliber, and the eventual development of some collection was excluded using US. If no more problems occurred, the drainage tube was removed progressively in the course of a few days.

From January 1, 1995, to January 31, 1997, 107 consecutive patients with HCCs at Tokyo University Hospital, Tokyo, Japan, were enrolled. They all fulfilled the selection criteria, were admitted to hospital, and underwent surgery. Types of liver dysfunction are shown in the following tabulation:

The one patient with HCC and primary biliary cirrhosis was considered noncirrhotic for the purpose of statistical evaluation, because the liver disease was at an early stage. Patient characteristics are presented in Table 1.

Portal venous branch embolization was performed for 2 patients who had 2 and 1 lesions in normal and fibrotic livers with maximum diameters of 10 and 9.5 cm, respectively. Major resections (1 sector or more) were performed in 34.6%, segmentectomy in 42.0%, and limited resection, including enucleation, in 23.4% (Table 2). Of the cirrhotic patients, 18.7%, 50.0%, and 31.3% underwent major resection, segmentectomy, and limited resection, respectively (Table 2).

A bilateral Student t test for unpaired data was used to compare the mean ICG15 rates and blood loss, and a χ² test for comparison of proportions was used to compare the proportions of patients with neoplastic vascular invasion and who required blood transfusion. The patients were grouped according to whether they had liver cirrhosis and/or the type of operation performed. Multiple logistic regression analysis and calculation of the odds ratio were performed to evaluate relative risk for morbidity for each of the factors considered, ie, age, sex, total serum bilirubin level, ICG15, total serum protein level, prothrombin time ratio, main tumor size, number of lesions, the presence or absence of portal branch and hepatic vein involvement, the coexistence or absence of liver cirrhosis, operative procedure, operation duration, blood loss, and blood transfusion. A D statistic was performed to validate the goodness of fit of the considered logistic model. Low D values and high P values meant the model fit well. The significance of each factor included in the analysis was evaluated using the Wald test, and the level of significance was set at P=.05.

The mean ICG15 value of the cirrhotic patients who underwent limited resections was 23.4%, whereas those of the cirrhotic patients who underwent segmentectomy and major resection were 14.7% (P=.02) and 14.0% (P=.02), respectively. The mean ICG15 values of the noncirrhotic patients who underwent limited resection, segmentectomy, and major resection were 21.9%, 13.1% (P=.04), and 12.4% (P=.009), respectively.

In the cirrhotic group, tumor thrombi were present in the portal branches or hepatic veins of 5 (25%) of 20 patients who underwent limited resection, 3 (9%) of 32 who underwent segmentectomy, and 5 (42%) of 12 who underwent major resection. Neoplastic invasion of intrahepatic vessels was evident in 1 (20%) of 5, 16 (64%) of 25, and none of 13 noncirrhotic patients who underwent limited resection, major resection, and segmentectomy, respectively. The incidence of HCC with vascular invasion was significantly higher in cirrhotic (P=.01) and noncirrhotic patients (P<.001) who underwent major resection than those who underwent segmentectomy.

Overall, the mean blood loss was 896.3 mL (SD, 698.9 mL; median, 687.5 mL; range, 40-4072 mL). The mean blood loss of the cirrhotic patients was 927.3 mL (SD, 755.2 mL; median, 663.5 mL; range, 40-4072 mL), whereas that of the noncirrhotic patients was 850.8 mL (SD, 604.5 mL; median, 704.5 mL; range, 180-3100 mL); these means did not differ significantly (P=.74). When both groups were stratified according to the type of operation, only the mean blood loss of the cirrhotic patients who underwent limited resection was significantly higher than that of the noncirrhotic patients who underwent the same procedure (P=.02) (Table 3).

Blood transfusions were administered to 10 (9.3%) of the 107 patients.

The overall mean operation duration was 394 minutes (SD, 127.1 minutes; median, 380 minutes; range, 115-870 minutes); for the cirrhotic patients, it was 385.9 minutes (SD, 137.5 minutes; median, 380 minutes; range, 115-870 minutes); and for the noncirrhotic patients, it was 406.7 minutes (SD, 108.8 minutes; median, 395 minutes; range, 200-705 minutes). There was no significant difference between the latter 2 mean values (P=.41). No significant differences were detected when the mean operation durations of the cirrhotic and noncirrhotic patients stratified according to the type of operation were compared (Table 4).

No 30-day mortality occurred, and all the patients returned to their normal daily lives.

The overall morbidity rate was 26.2% (Table 5). No major complications occurred, and no patient required reoperation. A chest tube was placed postoperatively in a patient with a pneumothorax; this treatment was successful, and the patient recovered completely.

The morbidity rate of patients with HCCs with associated cirrhosis was 29.7%, whereas that of the remaining noncirrhotic patients was 20.9%. Transient postoperative ascites and wound dehiscence only occurred in cirrhotic patients (Table 5).

The adopted model for performing multiple logistic regression analysis yielded a D value of 24.5 (P=.14), demonstrating good model fitting. The only factor associated with a significantly higher risk for postoperative morbidity was the operative procedure, as shown in Table 6. A shorter operation and lower prothrombin time ratio resulted in high odds ratios, but the effects of these factors did not reach statistical significance.

Although historically liver resection has been considered the treatment of choice for HCC, it is often unfeasible due to neoplastic features or, predominantly, impaired liver function. Most authors have reported a low resectability rate of the tumors (12%-28%) among patients referred to them,^{15 - 16} and, in many series, the morbidity and mortality rates related to the surgical procedures were very high, up to 47%¹⁷ and 23%,¹⁸ respectively. In particular, the coexistence of liver cirrhosis has a considerable adverse effect on the surgical results, with mortality rates of up to 37%,¹⁹ and for this reason, it is still considered a prohibitive condition for the surgical treatment of HCC. Consequently, different therapies have been proposed for managing HCC with associated liver cirrhosis. Although these new therapies initially enabled patients otherwise excluded from the surgical program to be treated, the good results obtained have generated confusion in the scientific community as to which treatment modalities should be chosen to manage HCCs and associated liver cirrhosis.^{20 - 21}

Despite the series composed of relatively small numbers of patients, orthotopic liver transplantation proved to be an appropriate treatment for HCCs occurring with liver cirrhosis and consequently could be the first choice.^{22 - 24} Unfortunately, its application is greatly limited due to the lack of donors. In fact, Mazzaferro et al²² reported recently that only 7% of the patients they observed had been treated with orthotopic liver transplantation. Moreover, the high morbidity and mortality associated with this approach and the recurrence rates that are not negligible, which are probably related to the difficulty of selecting suitable patients,²⁵ should be taken into consideration.

Other therapies, such as interstitial therapies^{26 - 29} and chemoembolization, have been devised. Some authors consider them not just adjuvant or palliative therapeutic modalities for surgical patients, but that they can be curative treatments.^{30 - 32} In particular, percutaneous ethanol injection (PEI) into small HCCs has resulted in long-term survival rates similar to those achieved with the best surgical series.³² The 5-year recurrence rates are also similar, ie, 57% to 100%^{33 - 37} for the surgical series and 64% to 100%^{32 ,38 - 40} with PEI. If we take into account the minimal discomfort to patients and the low cost of a procedure like PEI, which is performed on an outpatient basis, it seems reasonable to conclude that PEI is actually the most appropriate treatment for HCCs. However, a complete necrosis occurs in 40.8% to 73% of patients who undergo PEI,^{32 ,41} whereas appropriate anatomical liver resection, such as segmentectomy, guarantees good oncological radicality and improves patients' survival. In fact, patients who underwent limited resections, which, in terms of radicality, are comparable with PEI treatment that achieves a complete response, had significantly lower long-term survival rates^{5 ,42} and higher recurrence rates⁴² than those who underwent segmentectomies.

Thus, assuming that surgery is probably the more appropriate treatment from an oncological perspective, problems such as significant mortality and morbidity rates nevertheless limit its use in favor of more conservative modalities. The extreme care we take with each procedure involved in our preoperative, perioperative, and postoperative steps has enabled us to achieve, for the first time to our knowledge, a mortality rate below 2%¹⁰ and consequently, as reported above, no perioperative deaths.

Although a standard quantitative technique for evaluating the surgical risks for such patients has still not been defined and accepted worldwide, previous experience^{10 ,43} and the results of our prospective study demonstrate the adequacy of our decision tree for selecting cirrhotic patients with HCC for whom surgical resection is appropriate. In fact, multivariate analysis of the patients undergoing limited resection revealed they had a significantly higher risk for morbidity than those who underwent major resection or segmentectomy, despite the higher invasiveness of the latter procedures (Table 6). Moreover, Table 6 shows that the odds ratio of patients whose operations were shorter was higher than that of patients who had undergone longer and more aggressive liver resection procedures. These findings demonstrate that the adopted algorithm for patient selection correctly identified the patients at higher surgical risk for whom less invasive resections were suitable. Finally, the finding that cirrhotic patients were not at a significantly higher risk for postoperative morbidity than those who were not cirrhotic confirms the adequacy of this patient selection flow chart.

Measurement of the liver volume using preoperative computed tomography enables the resection volume to be estimated, and there was a significant correlation between the estimated liver volume and the weight of the resected specimen.¹³ This volume estimation in conjunction with our selection algorithm increases the accuracy of preoperative assessment of the surgical risk.

A period of bed rest during the preoperative hospital stay is recommended, as it seems to help the liver function of cirrhotic patients to recover.⁴⁴ To the same end, maintenance of a satisfactory nutritional status is mandatory, and it has been suggested, if necessary, that amino acid and lipid emulsions should be administered.⁴⁵ Ursodesoxycholic acid, Shosaiko-to,⁴⁶ and glycyrrhizin proved effective for reducing the serum alanine aminotransferase level, elevation of which seems to be a bad prognostic factor for patient outcome⁴⁷ and the risk for postoperative recurrence.³³

Portal venous branch embolization has become widely adopted due to its effectiveness in inducing remnant liver hypertrophy and reducing the volume of the hepatic portion to be resected.^{48 - 51} Recently, our group demonstrated how, in patients requiring major resection, portal venous branch embolization enables surgical procedures to be performed safely despite their low liver functional reserve (ICG15 values ranging from 10%-20%) or the necessity of removing more than 60% of the liver volume in patients with normal liver function.¹³

At laparotomy, J-shaped, upper-median, and inverted T-shaped incisions are generally preferred to the Mercedes star-shaped incision because a wider operative field can be obtained. The necessity for a thoracotomy is a subject of debate, but in our opinion, access to the chest is useful for the following reasons:

Certainly, a thoracophrenolaparotomy and a chest tube are more unpleasant for the patient, and postoperative complications such as atelectasis and pleural effusions may occur. Nevertheless, the opportunity to manage the liver better during a major resection involving the superior part of the right side of the liver widely justifies this approach.

The liver surgeon uses intraoperative US at every stage of the operation, and some surgical resections can be performed only if intraoperative US is available.^{54 - 55} Intraoperative ultrasound guidance enables anatomical liver resections with preservation of as much liver parenchyma as possible to be performed, and this is crucial for improving the radicality and safety of the surgical approach.

Minimization of intraoperative blood loss and the consequent lack of necessity for blood transfusion is a widely accepted goal,^{56 - 57} because these factors impair the patients' outcome⁵⁸ and prognosis.⁵⁹ Fan et al⁶⁰ recently published their experience with liver resections using the ultrasonic dissector to obtain a bloodless operative field; although they found the ultrasonic dissector was useful for reducing blood loss and that the procedure was not time-consuming, the transfusion rates with and without the ultrasonic dissector were relatively high (69.1% and 91.7%, respectively). Fewer than 10% of our patients with HCC treated surgically required blood transfusion. As a transfusion rate of up to 30% is generally considered acceptable,⁶¹ our findings demonstrate that very good results can be reached without using an ultrasonic dissector. The sparing of time and lower costs are other reasons to support the use of the Péan forceps dissection technique rather than the ultrasonic dissector.

At present, it generally is accepted that liver resection performed under intermittent warm ischemia is a safe and well-tolerated modality in patients with and without cirrhotic livers.⁶² In our experience of patients who have undergone limited resection, cirrhotic patients lose significantly more blood than noncirrhotic patients, suggesting that a method of reducing bleeding, such as warm ischemia, is particularly necessary in cirrhotic patients. Recently, Man et al⁶³ provided further confirmation of the usefulness of warm ischemia; their prospective randomized trial showed that the postoperative outcome of patients who underwent liver resection with the Pringle maneuver was better than that of those who underwent operation without it.

The goal of limiting blood transfusion can also be achieved by infusing plasma and a crystalloid to replace lost blood, thereby reducing the risk for postoperative symptomatic hemoconcentration and hyperbilirubinemia.⁵⁸ Minimization of back bleeding from accidentally injured hepatic veins is another way of reducing the need for blood transfusion, and reducing the inferior vena caval pressure by administering a muscle relaxant and reducing the tidal volume seems to be adequate to achieve this. Some authors have administered nitroglycerine to do so.⁶⁴

To achieve a perfect hemostasis, we always use fibrin, which is particularly effective for stopping accidental bleeding that can occur during the liver dissection from small holes on the walls of hepatic veins.

Some authors have suggested avoiding placement of a drain when performing liver surgery because they found it ineffective.^{65 - 66} However, there are very few reports about this specific issue in the field of liver surgery, and the experience from these studies is limited. Moreover, opinions are not always unanimous; some authors have actually affirmed the importance of drain placement.⁶⁷ Fortunately, postoperative massive bleeding is a rare occurrence, but it is an extremely alarming complication. The frequency of perihepatic infected fluid collections has been reported to be 2% to 20%,⁶⁶ biliary fistulas occurred in up to 8% of patients,^{66 ,68} and immunodeficiency and poor wound healing in cirrhotic patients affect these complications. As the rate of occurrence of these complications is significant, drain placement is mandatory because it plays a fundamental role in management.

With respect to postoperative care, some authors suggest achieving an energy support by administering lipid emulsions.^{45 ,69} However, we consider glucose administration provides enough energy, because oral nutrition should be restored as soon as possible after hepatectomy. We administer FFP postoperatively to replace blood loss and to keep the serum protein level within the normal range.

The policy we have proposed obviates most of the weak points of the surgical approach described herein, with no ensuing major complications and, moreover, no mortality. Although a reliable resectability rate could not be estimated because of the bias in patient selection (made at first by the referring physicians), the adopted selection algorithm guaranteed appropriate and safe treatment of all patients with HCCs admitted to our department who were oncologically suitable for a curative approach. As shown in our results, limited resections were performed in patients with high ICG15 values, whereas after staging the tumor, a choice between major resection and segmentectomy was made for patients whose ICG15 values permitted more radical procedures. The efficacy of our treatment is demonstrated by the long-term survival rates published previously.¹⁰

However, the goal of considering liver transplantation as the first-choice therapeutic option for HCC with associated liver cirrhosis seems hazardous, because the number of donors is limited, and the results in terms of perioperative mortality, morbidity, long-term hepatitis, and tumor recurrence are still not completely satisfactory. The results of chemoembolization are not yet well defined in terms of survival, but the ensuing morbidity and mortality are nevertheless significant.⁷⁰ Percutaneous ethanol injection seems to be the treatment of major interest because of its good long-term results, low cost, low morbidity, and almost no mortality.⁷¹ Moreover, its indications have been recently extended to include large HCCs, and, apparently, good results have been achieved.⁷² Despite these considerations, however, the real incidence of some problems, such as intraperitoneal seeding,^{71 ,73} may be underestimated. A recent Japanese survey of small, single, and well-differentiated HCCs revealed that the 5-year survival of patients who underwent surgery was significantly better than that of patients who underwent PEI (90% vs 72%; P<.05). Moreover, this survey revealed a significantly higher recurrence-free survival rate after resection than after PEI (48% vs 28%; P<.01).⁷⁴ Further confirmation of these data comes from a recent survey conducted by the Liver Cancer Study Group of Japan; 9326 patients with HCCs treated with liver resection had a better overall 4-year survival than 952 treated with PEI (55.1% vs 37.1%).⁴¹

We have demonstrated the safety of surgical resection for HCC in patients with liver cirrhosis. Our results and the oncological rationale indicate the surgical option is the best means of prolonging patients' survival. The good results obtained with an aggressive surgical approach for patients with advanced HCCs, even those with tumor thrombi in the portal trunk,⁷⁵ the therapeutic efficacy demonstrated by liver resection with respect to postoperative tumor recurrence,^{16 ,76} and, moreover, in patients with such recurrence, the feasibility and safety of aggressive resection with vascular reconstruction⁷⁷ are signs that with the advent of dedicated technologies and modalities such as intraoperative US and portal venous branch embolization, surgical treatment should be used more rather than only for restricted indications.

Furthermore, improved knowledge about the risk factors associated with HCC recurrences^{33 ,78} and their clonality,⁷⁹ together with the development of more effective therapies for their prevention,⁸⁰ will probably help with selection of surgical patients.

The surgical approach might be too dependent on the experience of the surgical team, thereby limiting its application. Certainly the surgeons performing liver surgery, particularly in patients with associated cirrhosis, must not only be well trained but also have extensive background knowledge and experience of hepatology and ultrasonography. Only with the full comprehension of this last concept will surgeons be able to perform safe liver surgery, which remains the treatment of choice for HCC, instead of shifting the indications to the ablation therapies.

This work was supported by a fellowship from the Japanese Foundation for Promotion of Cancer Research.

We thank Angela Simms and David Douglas for the English revision of the manuscript and Sugako Saitoh for her assistance in editing the article.

Reprints: Masatoshi Makuuchi, MD, PhD, HepatoBiliaryPancreatic Surgery Division, Department of Surgery, Graduate School of Medicine, University of Tokyo, 7-3-1 Hongo Bunkyo-ku, Tokyo 113-8655, Japan (e-mail: makuuchi-tky@umin.u-tokyo.ac.jp).